Support Vector Machines-Based Prediction of Elastic Modulus for Granite Rock

Abstract

AbstractThe modulus of elasticity (E) of the rock masses is an important parameter in the design of any kind of structure. The related data obtained from the in situ, as well as the laboratory of the modulus of elasticity, is extremely nonlinear. However, soft computing techniques are innovative techniques, which predict the desired output accurately where the data is extremely nonlinear. In the present study, a soft computing technique, such as support vector machines (SVMs), was used with polynomial and radial basis kernels. To obtain the desired prediction of modulus of elasticity (E) of the granite rock, index properties of the rock mass were used. The prediction using radial basis kernel is better than that of using polynomial kernel as evident from the coefficient of determination (R2). Sensitivity analysis of the study reveals that the porosity ‘n’ and Schmidt hammer rebound number ‘Rn’ were the two major input parameters influencing the prediction of the modulus of elasticity of the granite rock. The R2 of obtained models was compared with the previous studies, which reveals that the present study was predicting the modulus of elasticity of the granite rock better than the previous studies reported in the literature.KeywordsModulus of elasticityGranite rockSupport vector machinesPolynomial kernelRadial basis kernels